Tim Genewein of DeepMind Institute describes how scaling, new methods, agent coordination and recursive self-improvement could push artificial intelligence beyond human-level performance. He calls on researchers and policymakers to measure progress as systems approach superintelligence.




